KEY POINTS
- This single-center study from Stanford University developed and validated a simulation-free, single-session, MR-guided adaptive SABR workflow for primary and metastatic liver tumors. Dosimetric validation included 17 previously treated single-fraction plans from 16 patients, followed by initial prospective experience in five patients enrolled in a phase II clinical study (NCT07223307). The prospective protocol plans to recruit 62 patients to evaluate safety and one-year local control.
- The proposed workflow delivers 30–40 Gy in one fraction using an MRIdian MR-Linac, integrating treatment-day MRI, contour adaptation, plan reoptimization, quality assurance, and gated beam delivery. A diagnostic CT acquired within three months is used to establish reference-plan geometry and independently verify dose calculation, while a dedicated radiotherapy CT simulation visit is omitted. Patients perform reproducible breath-holds, verified using at least three 3D MRI acquisitions before online adaptation.
- The central technical innovation is a simplified bulk density override (BDO) method using three relative electron-density classes: body/soft tissue 1.0, air 0.0012, and lung 0.26. Additional density classes were introduced only when necessary. Among the 17 retrospective plans, an adipose override was required in three, and a separate liver override in one; the remaining cases used the three-class model without a dedicated bone-density assignment.
- BDO-based dose calculation closely reproduced CT-based dosimetry. Across target metrics, including GTV and PTV D98, D95, D50, D2, and mean dose, median differences were approximately zero and maximum differences were less than 2%. For chest wall endpoints, mean absolute dose differences were ≤0.07 Gy, with a maximum difference of 0.62 Gy and R² values of 0.997–1.000. Mean absolute differences for bowel, spinal cord, and liver-minus-GTV metrics were ≤0.05 Gy.
- Using stringent 1%/1 mm 3D gamma criteria, retrospective BDO-versus-CT agreement averaged 99.2% (range 96.4–100%), with no statistically significant differences in evaluated DVH endpoints. This agreement provides evidence that simplified density assignment can be sufficiently accurate for selected liver treatment geometries, although it does not independently validate all sources of MR-guided dose-calculation uncertainty.
- In the first five prospectively treated patients, four plans used the standard three-class model and one required an additional liver-density assignment. Maximum discrepancy in PTV V100% was 1.32 percentage points, while clinically relevant OAR dose metrics remained closely matched. Mean 1%/1 mm gamma passing rate was 98.8% (range 97.4–99.6%), and mean and median in-room treatment times were 94 and 86 minutes, respectively.
- Local control, toxicity, and patient-reported outcomes are not yet available. The ongoing phase II trial will evaluate one-year local control and early deterioration in albumin-bilirubin grade, among other endpoints. The current experience is limited to selected liver anatomies at one institution, and inaccurate lung–liver interface assignment, variable breath-hold reproducibility, or targets near gastrointestinal OARs could compromise safety; full physician and physics review remains necessary.
CLINICAL TAKEAWAY
Simulation-free, single-session MR-guided liver SABR appears technically feasible using simplified bulk-density assignment, with close agreement to conventional CT-based dose calculations. This approach could reduce treatment visits for selected patients, but evidence is currently limited to dosimetric validation and five treated cases, and its safety, local control, and operational advantages require prospective confirmation.